Embodiments of the invention relate generally to motor controls and, more particularly, to an apparatus and method determining a magnetizing curve of an induction machine.
High-performance sensorless vector control based on rotor flux orientation relies heavily on the parameter estimation of the induction machine. Many methods have been proposed for induction motor parameter identification, including online methods and the offline methods. Generally, online identifications of stator and rotor resistances are based on offline estimation values. A conventional method for rotor identification includes the locked rotor and no-load tests. However, the locked rotor and no-load tests cannot be realized in some engineering applications.
A method involving exerting different frequencies of single-phase sinusoidal excitations to the motor to imitate locked rotor and no-load situations has been proposed. A disadvantage of this method is that the frequency selection is very important, and the solution is very sensitive to sampling errors, which may result in a large rotor resistance identification error. Other methods have also been proposed based on statistical data, such as methods using a recursive least squares (RLS) algorithm. These methods can be highly precise but are often too demanding computationally, which leads to a long operation cycle.
Numerous applications use induction machines operating at a high speed far more than the rated speed. As the available voltage is limited under this high speed, the rotor flux reference are reduced as the speed increases, leading to operation in the flux weakening region. The issue of optimal magnetizing inductance updating in the flux weakening region has been a subject of great interest in recent decades. Online updating of magnetizing inductance can rely on the magnetizing curve identified offline. The magnetizing curve is typically made into a table that is stored in a controller, and the magnetizing inductance value is updated online by looking up the table according to the flux reference.
Common methods for magnetizing curve identification include curve fitting algorithms. In these methods, explicit functions are used (e.g., power functions, rational power functions, polynomials, etc.), and the coefficients of assumed functions are set accordingly. By selecting a group of magnetizing inductance estimation values under different magnetizing currents, the coefficients of the function can be solved. However, the real magnetizing curve relies on no definite function, and the magnetizing curve of one machine is different from that of another. So these methods may not be the best choice for accurate vector control.
It would therefore be desirable to have an apparatus and method for determining a magnetizing curve of an induction machine that overcome the aforementioned drawbacks.